A data-efficient intrusion detection framework tailored for Wireless Sensor Networks (WSNs) is proposed by leveraging active learning and metaheuristic optimization techniques. This framework addresses three major limitations of traditional models: data imbalance, inefficient hyperparameter tuning, and the need for large labeled datasets. To handle class imbalance, adaptive synthetic sampling generates synthetic instances for minority classes, particularly enhancing learning in complex regions of the feature space. For hyperparameter optimization, the Sandpiper Optimization (SO) algorithm is employed to fine-tune the regularization parameter of Logistic Regression (LR), leading to improved generalization. The issue of limited labeled data is tackled using Active Learning Uncertainty (ALU) and Entropy-based Active Learning (ALE), which query the most informative samples from the unlabeled pool, maximizing learning with minimal annotation effort. Simulation results show that LRALU, LRALE, and LRSO outperform traditional models with improvements of 18.18%, 19.48%, and 9.09% in accuracy; 9.30%, 1.16%, and 9.30% in precision; 18.18%, 19.48%, and 9.09% in recall; 12.20%, 8.54%, and 7.32% in F1-score; and 14.63%, 12.20%, and 9.76% in ROC-AUC, respectively. Additionally, log loss is reduced by 6.45%, 6.45%, and 35.48% for LRALU, LRALE, and LRSO, respectively. These results demonstrate that integrating intelligent sampling, active learning, and nature-inspired optimization significantly enhances intrusion detection performance in WSNs, providing an annotation-efficient solution for practical deployment.

The Internet of Things (IoT) has transformed modern technology by interconnecting physical devices to enable intelligent automation and real-time data exchange. However, securing IoT environments remains a critical challenge due to device heterogeneity, resource limitations, and vulnerabilities in lightweight communication protocols. Traditional Intrusion Detection Systems (IDS) often struggle with issues such as imbalanced datasets, suboptimal classification accuracy, difficulty in tuning hyperparameters, and a scarcity of labeled data. To address these limitations, we propose a novel IDS framework that integrates multiple advanced techniques. Initially, categorical labels are transformed using Label Encoding to facilitate effective model training. To mitigate data imbalance, the Localized Random Affine Shadowsampling (LoRAS) technique is applied, enhancing minority class representation. A Monte-Carlo Active Learning approach implemented on the DaNet architecture, termed MALD, is introduced to improve data efficiency by selectively querying the most informative samples. Additionally, we propose Elephant Herding Optimization applied to DaNet named EHODA to autonomously tune hyperparameters and maximize classification performance. Experimental results demonstrate that the proposed MALD and EHODA models significantly outperform conventional and state-of-the-art methods, achieving up to 93 % in Accuracy, Precision, Recall, and F1-Score, along with superior values in AUC-ROC of 0.98 and PR-AUC of 0.93. These findings affirm the effectiveness of our proposed framework for robust and adaptive intrusion detection in IoT environments.

Blockchain is a disruptive technology that has attracted considerable attention from scholars. The blockchain underlies cryptocurrencies and has rapidly expanded to other areas, including financial transactions and social media networks. However, concerns regarding the information security of social media users still exist regarding blockchain technology. The literature on blockchain online social media (BOSM) networks is growing rapidly because of their critical role in securing users’ information privacy and security. Cybersecurity remains a challenge faced by users on social media networks. Since the publication of BOSM, blockchain has become a widely discussed method for users’ information security. This comprehensive review identifies peer-reviewed articles on BOSM that underpin smart contracts, social media challenges, and research gaps. In this work, Kitchenham’s review guidelines are followed to conduct an in-depth review of the use of blockchain technology in the social media network literature published between January 2016 and March 2024, which reveals a significant increase in publications over the last eight years. A search of major academic databases, including Springer, ScienceDirect, ACM, IEEE Xplore, World Scientific, Taylor & Francis, and Wiley Online, yielded a final pool of 158 articles. The findings of the review indicate key insights concerning the techniques and applications of blockchain technology and challenges for the public via social media networks such as Twitter, Facebook, and Google+. This paper identifies important challenges such as deploying smart contracts, user information privacy, a lack of platform support, users’ reactions to blockchain technology, privacy protection and compensation, security system validation, online disinformation, scalability, and miscellaneous challenges to blockchain technology. Additionally, this review suggests several future research directions to improve the role of blockchain technology in overcoming the challenges of privacy, security, reliability, scalability, and trust in the area of social media networks.

ABSTRACT Hypertension continues to be a primary cause of global death, necessitating early and accurate forecasting for effective treatments. The existing methods have drawbacks such as class imbalance, poor modeling of sequential and spatial connections, high computation costs, and lack of interpretability, even though Deep Learning (DL) models offer possible solutions. To tackle these difficulties, we present VahigoNet, a novel blending DL model that incorporates vanilla recurrent neural networks (VRNN) for capturing temporal correlations, Google network for extracting hierarchical spatial features, and highway networks (HighwayNet) for adaptive feature refinements. To achieve strong generalization, we utilize the synthetic minority oversampling technique (SMOTE) for data balance. VahigoNet substantially outperforms baseline models, showing enhancements of 9.39% in accuracy, 10.27% in precision, 8.63% in recall, 9.39% in F1‐score, and 3.10% in area under the curve‐receiver operating characteristic. A 10‐fold cross validation method is utilized to assess the model's generalizability, markedly reducing overfitting and improving robustness. A paired t ‐test is performed to evaluate statistical significance, demonstrating that the enhancements are substantial and clinically relevant. Additionally, explainable artificial intelligence (AI) methodologies, including local Interpretable model‐agnostic explanations (LIME) and SHapley Additive exPlanations, are incorporated to provide both local and global perspectives on feature contributions. These explainability strategies enhance transparency, making VahigoNet a more interpretable and clinically reliable model for hypertension prediction. The results demonstrate that VahigoNet is an exceptionally efficient and transparent method, achieving a balance between predictive capability and practical relevance in medical diagnostics.